general-discussion > BASC in small brain region
Showing 1-4 of 4 posts
Display:
Results per page:
Jun 30, 2016  02:06 AM | Yuki Sakai
BASC in small brain region
Dear Bellec, Dear all,

 According to the explanation in NIAK web page (http://niak.simexp-lab.org/pipe_basc.htm...) and C-PAC web page (http://fcp-indi.github.io/docs/user/basc...), BASC is also useful to conduct clustering analysis and compare stability between groups even in small brain region (e.g. thalami or striatum).

 Although I cannot find the 'Manuel Garcia-Garcia, Brian Cheung, Adriana Di Martino, Pierre Bellec, Clare Kelly, F. Xavier Castellanos, Michael Milham. Stability of Functional Connectivity networks in the Basal Ganglia in ADHD. Poster. 2012' from the link on C-PAC web page (I guess it was available in the past), how can we optimize or validate the initial settings of some parameters in BASC below:
 * region growing threshold: In my understanding, this parameter have quite impact on the estimation of stability of clusters, especially when our region of interest is relatively small.
 * CBB: Is th length of temporal block (h) always the square root of the total number of time points (T) with some random deviation? Or do we have to set it up?
 * grid scales: How can we optimize or validate this parameter?

If possible, please share the poster itself or settings in 'Manuel Garcia-Garcia, Brian Cheung, Adriana Di Martino, Pierre Bellec, Clare Kelly, F. Xavier Castellanos, Michael Milham. Stability of Functional Connectivity networks in the Basal Ganglia in ADHD. Poster. 2012', or give me some suggestions.

Best,
 Yuki
Jul 5, 2016  05:07 PM | Pierre Bellec
RE: BASC in small brain region
Hi Yuki,

Sorry for the late answer. 

Re Manuel Garcia-Garcia's poster. I unfortunately do not have a copy readily accessible. There is a manuscript on that work which I am hopeful to see come out sooner rather than latter, at least in the form of a preprint. I will make sure to share it with you should this possibility materialize. 

Re choice of region growing size, it does indeed impact markedly stability estimates. For small regions you may be able to work at the voxel level. I would go with the number of regions: the pipeline runs in a reasonable time with up to 3000 regions. Beyond that it is starting to be really slow. I don't think the CPAC implementation features any dimension reduction prior to BASC, so you will have to work at the voxel level. 

Re choice of the number of clusters, the rationale behind MSTEPS applies as well to regions as it does to full brain analysis.  

Re the choice of the window length, it is based on the variability of the correlation coefficient between regional time series, wthich peaks at about sqrt( # time frames) -- that is described in a old relatively obscure paper (Bellec et al. Statistica Sinica 2008). This argument does not depend on the space of analysis (region vs brain). Note that the actual parameter is treated as a nuisance parameter and actually randomized in the NIAK BASC implementation (may be as well in CPAC). 

Best,

Pierre
Jul 17, 2016  05:07 AM | Yuki Sakai
RE: BASC in small brain region
Dear Bellec,

 Thank you for your kind reply.

 Re Manuel Garcia-Garcia's poster: I hope it would be published soon. Was it also conducted in voxel level manner or different settings?

 Re choice of region growing size: I also think that the voxel level analysis is reasonable, since the results through region growing seemed to be unstable. I run BASC on docker version of NIAK. How can I conduct voxel level BASC? According to some tutorials of NIAK, the region growing method is conducted before BASC. In addition, in the case of voxel level BASC, the brain template such as AAL will not be needed. How can I skip region growing and brain template?

 Re choice of the number of clusters: I understand that it is similar with whole brain analysis.

 Re the choice of the window length: I'm sorry for my ambiguous question. I know it doesn't depend on the volume of brain regions. I wonder how does the window length set up from number of time points (T) in docker version of NIAK (in which part of scripts?), and I have to set some parameters up for it or not.

Sincerely,
 Yuki
Jul 20, 2016  09:07 PM | Pierre Bellec
RE: BASC in small brain region
Dear Yuki,

The work by Manuel was implemented at the voxel level. 

For the region growing, I would think that if you use 

opt.thre_size = 0;

in the options of the BASC pipeline, the region growing will simply give one voxel per roi, so you will run a voxel level analysis. 

Make sure you specify a `files_in.mask` that is restricted to your area of interest. If you try to run BASC on the full brain at the voxel level you will saturate the memory (unless you have very big voxels).

If you don't want to use the AAL to reduce the memory load, specify your binary mask of interest in `files_in.areas`. See http://niak.simexp-lab.org/pipe_basc.html

Re the block length, check the help of niak_bootstrap_tseries
% BLOCK_LENGTH (if OPT.DGP == 'CBB')
% (integer, default [2*ceil(sqrt(T)) 3*ceil(sqrt(T))]) window width
% used in the circular block bootstrap. If multiple values are
% specified, a random parameter is selected in the list.

In other words, by default the bootstrap uses a block length of 2*the square root of the number of time samples, or 3* the square root of the number of time samples, and this is selected randomly for each bootstrap sample. You can change that parameter using

opt.stability_tseries.sampling.opt.block_length

but you will need to set a single value that applies to all you dataset (i.e. you cannot adjust the block length based on the number of time samples using that method). 

I hope this helps,

Pierre